mjhjmtu

I have a directory containing around 15K images, with x number of sequences

scene1_000000.png
scene1_000001.png
scene1_000002.png
scene1_000003.png
scene1_000004.png
scene1_000005.png
scene1_000006.png
scene1_000007.png
scene1_000008.png
 .
 .
 .
 .
scene2_000300.png
scene2_000000.png
scene2_000001.png
scene2_000002.png
scene2_000003.png
scene2_000004.png
scene2_000005.png
scene2_000006.png
scene2_000007.png
scene2_000008.png
 .
 .
 .
 .
scene2_000300.png
 .
 .
 .
 .
scene50_000000.png
scene50_000001.png
scene50_000002.png
scene50_000003.png
scene50_000004.png
scene50_000005.png
scene50_000006.png
scene50_000007.png
scene50_000008.png
 .
 .
 .
 .
scene50_000300.png

I want to keep from each sequence the first 150 sequences, and delete the rest.
So I will have for every scene, the sequences from 000000 to 000150

With a recent version of bash, you can use brace expansion for this:

rm scene*_000151..000300.png

Bash's brace expansion deals with leading 0s:

$ echo 000000..000005.png
000000.png 000001.png 000002.png 000003.png 000004.png 000005.png

So you can use 000151..000300 to generate the list of files you need.

Using find:

find . -maxdepth 1 -type f 
 -name 'scene*.png' 
 ! -name '*_0000[0-9][0-9].png' 
 ! -name '*_0001[0-4][0-9].png' 
 ! -name '*_000150.png' -print -delete

This would find all the files that you'd like to delete in the current directory (only).

The various -name flags do the following:

1. Only select the ones matching scene*.png.


2. Filter out (remove from selection) filenames that are in the range 000000.png to 000099.png.


3. Filter out filenames that are in the range 000100.png to 000149.png.


4. Filter out filenames that end in 000150.png.

... then delete the file if it's still considered. The 2nd, 3rd and 4th -name tests are negated to stop find from deleting those files (these are the ones we'd like to keep).

The names of the files deleted will also be printed before the actual deletion occurs.

I suggest that you run this with -delete removed first.

In zsh, the <m-n> construct matches strings that are numbers between m and n. Both are optional. Thus:

rm scene<->_<151->.png

(you could start the wildcard with scene*_ as well, but <-> would be a safeguard against * matching something unintended if there are files named according to a different pattern)

Maybe a little script :

echo '#!/bin/bash
# Test argument
if [ $# = 1 ]; then
 if [ -d $1 ]; then
 cd $1
 else
 echo "$1 is not a directory" >&2
 exit 1
 fi
fi

declare f # filname
declare -i fc # file counter
declare -i fcs # file conter for a scene
declare sn # scene number
declare csn # current scene number
declare -i dfc # delete file counter
declare -i pfc # preserve file counter

# Open logfiles
exec file-delete>>&3
exec file-preserve>>&4

for f in *; do
 ((fc++))
 sn=$f#scene
 sn=$sn%_*
 if [ "$sn" = "$csn" ]; then
 ((fcs++))
 else
 # New number scene
 fcs=1 
 csn=$sn
 dfc=0
 pfc=0
 echo
 fi
 if ((fcs > 150));then
 # Candidate for deleting
 echo "$f" >>&3
 ((dfc++))
 else
 # Preserve file
 echo "$f" >>&4
 ((pfc++))
 fi
 # Display in console
 echo -en "r scene $csn preserve:$pfc delete:$dfc"
done
echo

# Closing logfile
exec 3>&-
exec 4>&-

echo "Files scaned : $fc"
wc -l file-delete
wc -l file-preserve
' > ~/scanfile

For run it :

bash ~/scanfile [<pathdir>]

If is omitted, it works in current directory.

It does not delete anything, it writes two a file in scaned directory (file-delete and file-preserve).

After the check, the files to be deleted can be moved with :

[cd <pathdir>]
mkdir tmp
while read; do mv -v $REPLY tmp; done < file-delete

If the result is good, then

rm -rf tmp
rm file-delete file-preserve